[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

EP2207348A2 - Empfehlungsverfahren und System zur Cross-Domain-Empfehlung - Google Patents

Empfehlungsverfahren und System zur Cross-Domain-Empfehlung Download PDF

Info

Publication number
EP2207348A2
EP2207348A2 EP09150336A EP09150336A EP2207348A2 EP 2207348 A2 EP2207348 A2 EP 2207348A2 EP 09150336 A EP09150336 A EP 09150336A EP 09150336 A EP09150336 A EP 09150336A EP 2207348 A2 EP2207348 A2 EP 2207348A2
Authority
EP
European Patent Office
Prior art keywords
domain
feature
new
items
information
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
EP09150336A
Other languages
English (en)
French (fr)
Other versions
EP2207348A3 (de
Inventor
Mauro Barbieri
Serverius Pronk
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Funke Digital Tv Guide GmbH
Original Assignee
Axel Springer Digital TV Guide GmbH
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Axel Springer Digital TV Guide GmbH filed Critical Axel Springer Digital TV Guide GmbH
Priority to EP09150336A priority Critical patent/EP2207348A3/de
Publication of EP2207348A2 publication Critical patent/EP2207348A2/de
Publication of EP2207348A3 publication Critical patent/EP2207348A3/de
Ceased legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
    • H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
    • H04N21/44213—Monitoring of end-user related data
    • H04N21/44222—Analytics of user selections, e.g. selection of programs or purchase activity
    • H04N21/44224—Monitoring of user activity on external systems, e.g. Internet browsing
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N5/00—Details of television systems
    • H04N5/44—Receiver circuitry for the reception of television signals according to analogue transmission standards
    • H04N5/445—Receiver circuitry for the reception of television signals according to analogue transmission standards for displaying additional information
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
    • H04N21/4508—Management of client data or end-user data
    • H04N21/4532—Management of client data or end-user data involving end-user characteristics, e.g. viewer profile, preferences
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
    • H04N21/462—Content or additional data management, e.g. creating a master electronic program guide from data received from the Internet and a Head-end, controlling the complexity of a video stream by scaling the resolution or bit-rate based on the client capabilities
    • H04N21/4622—Retrieving content or additional data from different sources, e.g. from a broadcast channel and the Internet
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
    • H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
    • H04N21/4667—Processing of monitored end-user data, e.g. trend analysis based on the log file of viewer selections
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
    • H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
    • H04N21/4668—Learning process for intelligent management, e.g. learning user preferences for recommending movies for recommending content, e.g. movies
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
    • H04N21/47—End-user applications
    • H04N21/475—End-user interface for inputting end-user data, e.g. personal identification number [PIN], preference data
    • H04N21/4756—End-user interface for inputting end-user data, e.g. personal identification number [PIN], preference data for rating content, e.g. scoring a recommended movie

Definitions

  • the present invention relates to an apparatus, a method and a computer program product for providing recommendations for content items, e.g. a product or service, associated with a new domain to a user using available profile information for content items associated with known domains.
  • content items e.g. a product or service
  • IPTV Internet Protocol TV
  • P2P Peer-to-Peer networks
  • Mobile TV etc.
  • convergence may be seen as the product of the transition from analogue to digital, according to which all forms of content, e.g. voice, text, pictures, audio and video, are undifferentiated bits of data.
  • EPG Electronic Programme Guide
  • recommender technology is steadily being introduced into the market.
  • websites offer some kind of recommender functionality to support users in finding content items, such as movies or books.
  • electronics devices e.g. personal video recorders (PVRs)
  • PVRs personal video recorders
  • recommenders are becoming a popular tool to retrieve, from a vast amount of content items such as from repositories for audio/video (A/V) content, e.g. songs, television (TV) programs, movies, etc., or from catalogues for products, e.g. books, and the like, only those items a user or a group of users likes.
  • Recommenders can be offered as a stand-alone service, or as add-on functionality to an existing service. They increasingly appear in consumer devices, such as digital video recorders (DVRs).
  • DVRs digital video recorders
  • recommender systems are increasingly being applied to individualize or personalize services and products by learning a so-called user profile that may be based on rating feedback from individual or a group of users on selected content items.
  • machine-learning techniques can be used to infer the ratings for new items.
  • a recommender can typically be configured to learn the preference of a user, based on ratings that the user supplies on items; where ratings may be simple binary classifications, such as "like” and "dislike", respectively, or based on a more elaborate classification into a range of like-degrees.
  • the user rating provides an interface by means of which the user can teach the recommender about his preferences.
  • there are two types of recommender systems those based on a community of users and those based on metadata.
  • the first type of recommender systems is known as so-called collaborative filtering, where either (i) members of the community are characterized by the ratings they give to items or (ii) items are characterized by the ratings they receive from the members of the community. These characterizations (i) and (ii) may be used to define similarities among users or items, respectively. Thus, for a specific member of the community and a specific item that has not yet been rated by this member, these similarities can be used to infer for this member a rating for this item by combining ratings of similar users or similar items, respectively.
  • collaborative filtering can be seen as the ability to rate and/or tag videos or audio tunes. The level to which systems may use collaborative filtering ranges from simply rating content up to building "profiles" to provide recommendations and thus content to users.
  • the second type of recommender systems uses available metadata about items, which typically comes in the form of features (or attributes) and therewith associated values or lists of values.
  • the rating history of a user can be exploited to build a profile of this user, geared towards a particular machine-learning algorithm. For instance, for naive Bayesian classification, a popular and simple machine learning algorithm, this profile can be thought of as containing like-degrees for individual feature-value pairs.
  • the second type of recommender systems only requires the likes and dislikes of a single user, resulting in several advantages over the first type in that there is no need to create a community of users, no need for some central communication or processing facility, and in that privacy is thus much less of an issue.
  • an advantage of the first type over the second type of recommender systems is that no metadata is required, making collaborative filtering inherently suitable to provide recommendations across multiple domains, that is, to consider books, movies, food items, in one context.
  • the gist of the invention is using rating histories of one or more known domains to produce or infer a rating history for content items in a new domain.
  • this is achieved by means of computing like-degrees for feature-value pairs for the new domain based on forming and using translations and/or relations between information from profiles in the known domains and in the new domain. These translations or relations are exploited to extend profile information, comprised of like degrees of feature-value pairs, in the known domains into the new domain. Moreover, thereby inferred new profile information for new domains may be subsequently generalized.
  • the method underlying the invention may be separated into two aspects:
  • the first aspect concerns how to extend profiles from the known domain(s) to the new domain using translations and relations between feature spaces of the known domain(s) and the new domain.
  • the second aspect concerns the generalization of the so obtained profile information for content items in the new domain by using recommendation techniques.
  • a rating history may be understood as a list of content items of the respective domain, each connected with a classification, which may be e.g. a rating provided by or monitoring a user. This is especially advantageous since not all machine learning algorithms use like-degrees, but, in general, make use of rating histories, which each algorithm for itself translates/digests into a profile.
  • Such rating history can be transformed into profile information that can be geared towards a machine-learning algorithm.
  • profiles are merely positive and negative counts for individual feature-value pairs, from which like-degrees can be inferred. But for decision trees, neural networks, or support vector machines, the profile is less intuitive, and not necessarily can like-degrees be inferred.
  • rating histories generally a common characteristic is the use of rating histories.
  • the cross-domain recommender of the invention such useful rating history can be generated in a new domain. Accordingly, the whole cross-domain recommendation functionality is preferably based on profiles based on like degrees of feature-value pairs.
  • the apparatus for controlling a recommender system to provide cross-domain recommendations comprises:
  • the apparatus further comprises a transformation unit which is configured to transform available information in form of rating histories in the known domains into the respective profile information for the respective domain.
  • a rating history in this context, comprises content items of the domain each in connection with a respective classification of the content item. It goes without saying that there may be a translation unit for each known domain or one or several units for that purpose.
  • the translation information is stored as at least one translation table defining the linking information.
  • the extension unit is configured to perform the extension operation by inferring like-degrees of feature-value pairs in the new domain based on like-degrees of corresponding feature-value pairs in the at least one known domain in accordance with the translation information.
  • the extension unit is configured to calculate for a given feature-value pair in the new domain, a weighted, normalized average of the like-degrees of feature-value pairs in the at least one known domain that are linked by the associated translation table to the given feature-value pair in the new domain, and to use the calculated weighted, normalized average as a like-degree for the given feature-value pair in the new domain.
  • the extension unit is further configured to set the weights based on confidence values supplied by or learned by monitoring the user.
  • the confidence values may reflect the confidence the user of the system or the system itself has in the like-degree of each of the feature-value pairs.
  • the extension unit may be further configured to set the weights based on the respective like-degrees themselves.
  • Confidence values may be calculated from a rating history, as can be gathered e.g. from V. Pronk et al., in "Incorporating confidence in a naive Bayesian classifier", LNAI 3538, Springer, Proceedings of the 10th International Conference on User Modeling, UM'05, Edinburgh, Scotland, July 24-29, 317-326 .
  • the extension unit can be further configured to further extend the known profile information of at least one known domain as profile information for content items of the at least one new domain based on relation information between features in the new domain and/or features in the known domain.
  • the relation information comprises an ontology or, if the features are keywords, a thesaurus.
  • the extension unit is further configured to combine the results of multiple extension operations based on information from multiple known domains.
  • the recommender unit is further configured to individually classify the items in the chosen set of items using their computed scores by at least one classification value from a predetermined group of classification values, thereby providing a new rating history for the new domain.
  • the new rating history generated by the cross-domain recommender for the new domain can be used by any machine learning algorithm to generate a "full" profile, i.e. corresponding to the specifics of the algorithm.
  • the "full" profile is an extension of the profile initially used to create a rating history in the new domain.
  • the cross-domain recommender of the invention includes generating a rating history in the new domain.
  • the invention further relates to a method of controlling a recommender system providing cross-domain recommendations.
  • the method comprises the same advantages as the afore-described apparatus.
  • said method comprises the steps: a) storing known profile information for content items, defined by at least one feature and belonging to at least one known domain, wherein the profile information further comprises like-degrees for at least one feature-value pair for at least one user of the apparatus; b) storing profile information for content items, defined by at least one feature and belonging to a new domain; c) forming translations, which define linking information between at least one feature of at least one known domain and at least one feature of the new domain, or relations, which define linking information between features in the new domain and/or features in the at least one known domain, respectively; and d) extending the known profile information as new profile information for the content items of the new domain based on the translation information, and f) generalizing the generated profile in the new domain by i) using a subset of features of the new domain for which information is available, ii) choosing a set of items in the new domain, preferably randomly, and iii) computing, using the subset of features, the scores for each of the items in the apparatus
  • the method may further comprise the step: g) classifying the items in the chosen set of items using their computed scores by at least one classification value from a predetermined group of classification values, thereby providing a new rating history in the new domain.
  • the invention also relates to a computer program product that comprises code means or coded instructions, which cause a computer device or processor to produce or perform the steps of one of the methods of the invention, when the code means or coded instructions are executed on the computer device or processor.
  • code means or coded instructions which cause a computer device or processor to produce or perform the steps of one of the methods of the invention, when the code means or coded instructions are executed on the computer device or processor.
  • Embodiments of the present invention will now be described based on an exemplary recommender system which is able to generate ratings and thus recommendations on new content items in a new domain, such as books, TV programs, movies, etc., based on information about content items in at least one known domain.
  • a new domain such as books, TV programs, movies, etc.
  • Fig. 1 shows a schematic block diagram of the recommender system which comprises an information data store 103 connected to at least one source (S) 101, which may be, for example, A/V content repositories, product catalogues, and the like.
  • an A/V content repository can be an electronic program guide (EPG) service provided via the Internet, which provides information data on content items, for example, on television (TV) programs.
  • EPG electronic program guide
  • the information data store 103 can be connected to at least one filter (F) 105, which is associated with a personalized content channel. It is noted that any number of personalized content channels may be provided.
  • each personalized content channel may have an own recommender engine 107 associated therewith.
  • Each recommender engine 107 and hence personalized content channel has a profile (P) 109 based on a rating history (RH) 110 associated therewith.
  • the output of the recommender engine 107 is connected to a scheduler (SCH) 111.
  • the scheduler 111 is connected to a storage device 113, e.g. one or more hard disk drives of, for example a DVT or PVR or the like, and to a selector (SEL) 115.
  • the information data store 103 is also connected to a content source (CS) 117.
  • the content source 117 provides, for example, A/V content items in a broadcasting or on-demand fashion.
  • the content source 117 may provide besides the content information additional information data, i.e. metadata, such as the EPG information inside the video or audio signal.
  • the content source 117 is further connected to the selector 115 comprising at least one set of content isolation means, e.g. a tuner or the like, allowing to isolate one or more content items for recording on the storage device 113.
  • the output of the selector 115 is connected to the storage device 113.
  • Information data or metadata for a current content item to be played out on one personalized content channel is received from the source 101, i.e. from, for example, the Internet.
  • information data or metadata may be obtained via other means, e.g. via transmission in the vertical blanking interval of an analogue TV broadcast signal or contained within digital video broadcast (DVB) transport streams, or any combinations of any of the above.
  • a content item may be a TV program, data stream containing video and/or audio data or a segment of a program etc.
  • the information data may be associated with multiple domains, such as “TV shows”, “movies”, “books”, “food”, etc.
  • content items in each domain may be characterized by one or a plurality of features (or attributes).
  • feature values can be associated with a content item, e.g. for the domain "movies”, there may be features such as “title”, “actors”, “director” and “genre”.
  • the features and the associated feature-values represent metadata for the respective content items, where each feature-value pair (or attribute-value pair) can further be associated with like-degrees in accordance with the individual rating of a user.
  • each profile 109 is based on the information data together with data indicating the "likes” and "dislikes" of the respective user.
  • the rating in the simple form of e.g. "like” and “dislike” can be based, for instance, on user feedback on content items that pass the associated filter 105. This feedback can be given as explicit rating by the users that use the particular personalized content channel. Ratings by the user can be made in several ways. For example, the user can, e.g. by means of a remote control device as user interface (UI), indicate for a currently selected content item or a given feature of the current content item his or her rating, i.e. for example "like” or “dislike", by pressing appropriate buttons on the user interface, whilst viewing the current content item. In a more advanced setting, a "like” degree on a discrete or continuous scale can be provided or calculated instead of just a "like” or “dislike” classification.
  • UI user interface
  • use or user profiles can be derived using implicit profiling and explicit profiling.
  • Implicit profiling methods derive content use profiles unobtrusively from the user's use histories, e.g., sets of TV shows watched and not watched.
  • Explicit profiling methods may derive content use profiles by letting the user specify ratings on the level of content items.
  • this information data is forwarded to the recommender engine 107.
  • the recommender engine calculates a score, based on its associated profile 109, for this subsequent content item.
  • the information data associated to the subsequent content item is then forwarded, along with the computed rating, to the scheduler 111, which subsequently computes e.g. a recording schedule that will be used to schedule the recording of content items offered by the recommender engine 107 onto the storage device 113.
  • the scheduler 111 may primarily consider the content items of high score or rating while still considering sufficient new content for each personalized content channel.
  • the recording schedule computed by the scheduler 111 can be used to instruct the selector 115 to select the content items available from the content source 117 to record those selected content items on the storage device 113.
  • a new domain N can be characterized by features and its associated values, some of which, due to the above mentioned correlation, correspond directly or closely to some features and associated values in the known domain K.
  • the feature “genre” in the domain of “movies” has a strong relation to the feature “genre” in the domain of "books”.
  • the genre “science fiction” occurs in both domains “movies” and “books”.
  • the genre “romantic fiction” in the domain “books” translates into the two genres "romance” and "fiction” existing in the domain “movies”.
  • a possibly handcrafted, i.e. predetermined, or generated by machine learning a translation operation can be performed by means of e.g. a translation table between the two feature spaces.
  • a translation table between the two feature spaces.
  • a weighted, normalized average of the like-degrees of those feature-value pairs in the known domain K, which translate into the given feature-value pair in new domain N is calculated and used as a like-degree for the latter.
  • the weights are based on the confidence the user has in each of the involved feature-value pairs.
  • the weights are based on the respective like-degrees themselves. Any weighting that is carried out can also be based on the maturity of the profiles in the respective, known domain K.
  • relations between feature-value pairs in the known domain K and/or in the new domain N are used.
  • Such relations can be implemented in or defined by an ontology.
  • an ontology is considered as a formal representation of a set of features within a domain and the relationships between those features. Accordingly, the ontology can be used to reason about the properties of that domain, as well as may be used to define the domain. In simple cases, where a domain is defined simply by keywords, such relations can be implemented as a thesaurus.
  • the feature “genre” in the domain “movies” may, as described above, borrow from the domain “books”. Further, the feature “country of publication” in the domain “movies” may borrow from both the domains “books” and “travel”. Thus, more information is provided with respect to the feature "country of publication” than when only borrowing from one known domain Ki.
  • the feature "country of publication” in the domain “movies” can be inferred from the feature "location" of the domain “news items” by using the ontology mentioned above.
  • a generalization operation is performed by making use of a recommender that only takes those features into account for which there is already information available.
  • a set of content items in the new domain N is chosen, preferably randomly, and the set should not be too small.
  • the recommender using the subset of features, computes a score for each of these items and classifies each of them, for example as "positive", “negative", or, optionally, "undetermined”.
  • the "positives” and “negatives” can next be used as input, i.e. as a new rating history, to train a recommender that takes all features into account.
  • the foregoing described method and system for cross-domain recommendation can be used for online setting.
  • a user rates a content item in a given domain, this will result in an update of his/her information on like degrees of feature-value pairs in this domain.
  • this update can propagate to different, other domains connected directly or indirectly to the given domain by the herein described translation tables, ontologies, or thesauri.
  • Fig. 2 shows a schematic diagram of basic steps, elements or components of a control apparatus, which implements the proposed cross-domain recommendation functionality. Given a new content item out of a new domain N, 304, for which domain no profile information comprised of like degrees of feature-value pairs is available, and which item may be offered to the user.
  • the control apparatus or system is configured to use profile information comprised of like degrees of feature-value pairs contained in known domains K.
  • Such profile information in the known domains may be gathered or derived from rating histories available in the known domains K, 300.
  • a transformation unit 301 the rating information in the respective rating histories can be transformed or translated into profile information 302 comprised of like degrees of feature-value pairs for items of the respective known domain K.
  • At least one predetermined translation operation is implemented by an extension unit 320 based on information provided by in a translation table 310.
  • the information in the translation table 310 links the feature space of the new domain N, 304, with the feature space of a known domain K, 300, and thereby, with the respective profile information, comprised of like degrees of feature-value pairs, 302, for the user in the respective known domain 300.
  • the information defined in the translation table 310 is provided to the extension unit 320.
  • the extension unit 320 can infer like-degrees of feature-value pairs 306 in the new domain N, 304, based on the known like-degrees of corresponding feature-value pairs 302 in the respective known domain K, 300. This results in the generation of profile information, i.e. in the form of like degrees of feature-value pairs 306 for content items of the new domain N, 304.
  • the extension unit 320 may be configured to calculate a weighted, normalized average of the like-degrees of a feature-value pair 302 in a known domain K, 300, which translate by means of the translation table 310 into the given feature-value pair in new domain N, 304.
  • the weights can also be fed into the extension unit 320 as weight values that are based on the confidence values that have been associated with each of the involved feature-value pairs in the known domain(s).
  • the extension unit 320 can be configured to base the weights on the respective like-degrees themselves.
  • a recommender unit 330 performs a generalization operation on the profile information comprised of like degrees of feature-value pairs 306 gathered for the new domain N, 304, by the extension unit 320.
  • the recommender unit 330 is configured to use only those features in the new domain N, 304, for which already information is available as result of the extension operation.
  • the recommender unit 330 is configured to select a set of content items, preferably by means of a random generator, in the new domain N, 304.
  • the recommender unit 330 is further configured to compute scores for each of these content items by using the subset of features. Based on the computed like-degrees, the recommender unit 330 is configured to classify each of the content items, e.g. as "positive", "negative", or, optionally, "undetermined”.
  • Recommender 330 could be a naive Bayesian classifier, which produces a rating history in the new domain, to be used by a suitable machine-learning algorithm in that domain.
  • the list of content items in connection with its respective classifications i.e. in the example the values "positives” and “negatives”
  • any known machine-learning algorithm may be used to realize the further recommender unit.
  • naive Bayesian classification For example, learning algorithms that can make use of rating histories, such as naive Bayesian classification, decision trees, support vector machines, neural networks etc., may be applied. Each algorithm for itself translates/digests a rating history into a profile. As mentioned above, for naive Bayes, a profile is merely a collection of "positive” and “negative” counts for individual feature-value pairs, from which like-degrees can be inferred.
  • the cross-domain recommendation functionality of the invention can be applied to (Internet-enabled) TV sets, e.g. IPTV, PVRs, set-top boxes, audio systems including portable audio, and services including Internet video and music services, mobile phones, personal digital assistants (PDAs), personal computers (PCs) and all devices where recommenders are used to collect, filter, and present content from multiple sources to their users.
  • IPTV Internet-enabled
  • PVRs personal digital assistants
  • PCs personal computers
  • the invention is thus not restricted to recommenders of television or film content, but can be applied to music, theatre shows, books and all types of products and services for which recommenders can be built.
  • an apparatus, method, and computer program product for controlling a recommender system have been described.
  • the ability to make cross-domain recommendations is limited by the scope of the metadata in a profile.
  • a profile concerning movies cannot be used directly to do recommendations about food items.
  • the problem is that a user who has not yet built a profile in a certain domain cannot receive good recommendations in that domain.
  • An apparatus for controlling a recommender system is disclosed which provides recommendations in a new domain by using one or more profiles from other, known domains. This is achieved by forming or using translations or relations between the known domains and the new domain and by exploiting these translations or relations to extend the profiles in the known domains into the new domain. The thus generated profile in the new domain may subsequently be generalized.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Social Psychology (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Finance (AREA)
  • Development Economics (AREA)
  • Strategic Management (AREA)
  • Accounting & Taxation (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • General Physics & Mathematics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Human Computer Interaction (AREA)
  • Apparatus For Radiation Diagnosis (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
EP09150336A 2009-01-08 2009-01-09 Empfehlungsverfahren und System zur Cross-Domain-Empfehlung Ceased EP2207348A3 (de)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP09150336A EP2207348A3 (de) 2009-01-08 2009-01-09 Empfehlungsverfahren und System zur Cross-Domain-Empfehlung

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
EP09150225 2009-01-08
EP09150336A EP2207348A3 (de) 2009-01-08 2009-01-09 Empfehlungsverfahren und System zur Cross-Domain-Empfehlung

Publications (2)

Publication Number Publication Date
EP2207348A2 true EP2207348A2 (de) 2010-07-14
EP2207348A3 EP2207348A3 (de) 2013-02-13

Family

ID=41228707

Family Applications (1)

Application Number Title Priority Date Filing Date
EP09150336A Ceased EP2207348A3 (de) 2009-01-08 2009-01-09 Empfehlungsverfahren und System zur Cross-Domain-Empfehlung

Country Status (1)

Country Link
EP (1) EP2207348A3 (de)

Cited By (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013130834A1 (en) * 2012-03-01 2013-09-06 Qloo, Inc. Personalized cross-domain recommender system
EP2760215A1 (de) * 2013-01-25 2014-07-30 Samsung Electronics Co., Ltd Bildanzeigevorrichtung, Verfahren für Inhaltsempfehlungsinformationsempfang, Server und Verfahren zur Inhaltsempfehlung
US20140279756A1 (en) * 2013-03-18 2014-09-18 The Echo Nest Corporation Cross media recommendation
US8909583B2 (en) 2011-09-28 2014-12-09 Nara Logics, Inc. Systems and methods for providing recommendations based on collaborative and/or content-based nodal interrelationships
US9009088B2 (en) 2011-09-28 2015-04-14 Nara Logics, Inc. Apparatus and method for providing harmonized recommendations based on an integrated user profile
WO2015054361A1 (en) * 2013-10-08 2015-04-16 The Echo Nest Corporation Systems, methods, and computer program products for providing contextually-aware video recommendation
US9392308B2 (en) 2011-08-04 2016-07-12 Thomson Licensing Content recommendation based on user location and available devices
WO2016207308A1 (en) * 2015-06-26 2016-12-29 British Telecommunications Public Limited Company Digital content provision
EP3203756A1 (de) * 2016-02-02 2017-08-09 Thomson Licensing Verfahren zur bestimmung eines personalisierten profils und zum filtern von auszügen eines multimediainhalts, zugehörige vorrichtungen, computerprogrammprodukt und computerlesbares trägermedium
US9798823B2 (en) 2015-11-17 2017-10-24 Spotify Ab System, methods and computer products for determining affinity to a content creator
US9882790B2 (en) 2012-08-23 2018-01-30 Teknologian Tutkimuskeskus Vtt Method and apparatus for a recommendation system based on token exchange
US10339586B1 (en) * 2013-08-23 2019-07-02 Amazon Technologies, Inc. Techniques for identifying similar products
CN110110209A (zh) * 2018-01-22 2019-08-09 青岛科技大学 一种基于局部加权线性回归模型的交叉推荐方法及系统
US10467677B2 (en) 2011-09-28 2019-11-05 Nara Logics, Inc. Systems and methods for providing recommendations based on collaborative and/or content-based nodal interrelationships
US10516906B2 (en) 2015-09-18 2019-12-24 Spotify Ab Systems, methods, and computer products for recommending media suitable for a designated style of use
CN111046280A (zh) * 2019-12-02 2020-04-21 哈尔滨工程大学 一种应用fm的跨领域推荐方法
WO2020096272A1 (en) * 2018-11-05 2020-05-14 Samsung Electronics Co., Ltd. System and method for cross-domain recommendations
US10789526B2 (en) 2012-03-09 2020-09-29 Nara Logics, Inc. Method, system, and non-transitory computer-readable medium for constructing and applying synaptic networks
US10860646B2 (en) 2016-08-18 2020-12-08 Spotify Ab Systems, methods, and computer-readable products for track selection
WO2021176716A1 (ja) * 2020-03-06 2021-09-10 日本電気株式会社 嗜好推定装置、嗜好推定方法および嗜好推定プログラム
US11151617B2 (en) 2012-03-09 2021-10-19 Nara Logics, Inc. Systems and methods for providing recommendations based on collaborative and/or content-based nodal interrelationships
EP3902277A4 (de) * 2019-01-07 2022-06-08 Samsung Electronics Co., Ltd. Verfahren zur bereitstellung einer liste mit empfohlenen inhalten und entsprechende elektronische vorrichtung
US11379513B2 (en) 2019-08-30 2022-07-05 Spotify Ab Systems and methods for providing content based on consumption in a distinct domain
GB2606878A (en) * 2018-05-25 2022-11-23 Thinkanalytics Ltd Content recommendation system
US11727249B2 (en) 2011-09-28 2023-08-15 Nara Logics, Inc. Methods for constructing and applying synaptic networks
US11812107B2 (en) 2018-05-25 2023-11-07 Thinkanalytics Ltd Content recommendation system and method
US12114030B1 (en) 2023-04-04 2024-10-08 Thinkanalytics Ltd. Managing feature distribution

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
PRONK, V. ET AL.: "Incorporating user control into Recommender Systems based on Naive Bayesian Classification", PROC. ACM RECOMMENDER SYSTEMS, RECSYS'07, MINNEAPOLIS, MN, USA
V. PRONK ET AL.: "Proceedings of the 10th International Conference on User Modeling, UM'05, Edinburgh", 24 July 2007, SPRINGER, article "Incorporating confidence in a naive Bayesian classifier", pages: 317 - 326

Cited By (47)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9392308B2 (en) 2011-08-04 2016-07-12 Thomson Licensing Content recommendation based on user location and available devices
US9449336B2 (en) 2011-09-28 2016-09-20 Nara Logics, Inc. Apparatus and method for providing harmonized recommendations based on an integrated user profile
US8909583B2 (en) 2011-09-28 2014-12-09 Nara Logics, Inc. Systems and methods for providing recommendations based on collaborative and/or content-based nodal interrelationships
US9009088B2 (en) 2011-09-28 2015-04-14 Nara Logics, Inc. Apparatus and method for providing harmonized recommendations based on an integrated user profile
US10467677B2 (en) 2011-09-28 2019-11-05 Nara Logics, Inc. Systems and methods for providing recommendations based on collaborative and/or content-based nodal interrelationships
US11651412B2 (en) 2011-09-28 2023-05-16 Nara Logics, Inc. Systems and methods for providing recommendations based on collaborative and/or content-based nodal interrelationships
US10423880B2 (en) 2011-09-28 2019-09-24 Nara Logics, Inc. Systems and methods for providing recommendations based on collaborative and/or content-based nodal interrelationships
US11727249B2 (en) 2011-09-28 2023-08-15 Nara Logics, Inc. Methods for constructing and applying synaptic networks
US20140074650A1 (en) * 2012-03-01 2014-03-13 Qloo, Inc. Personalized cross-domain recommender system
WO2013130834A1 (en) * 2012-03-01 2013-09-06 Qloo, Inc. Personalized cross-domain recommender system
US10789526B2 (en) 2012-03-09 2020-09-29 Nara Logics, Inc. Method, system, and non-transitory computer-readable medium for constructing and applying synaptic networks
US11151617B2 (en) 2012-03-09 2021-10-19 Nara Logics, Inc. Systems and methods for providing recommendations based on collaborative and/or content-based nodal interrelationships
US9882790B2 (en) 2012-08-23 2018-01-30 Teknologian Tutkimuskeskus Vtt Method and apparatus for a recommendation system based on token exchange
EP2760215A1 (de) * 2013-01-25 2014-07-30 Samsung Electronics Co., Ltd Bildanzeigevorrichtung, Verfahren für Inhaltsempfehlungsinformationsempfang, Server und Verfahren zur Inhaltsempfehlung
US9613118B2 (en) 2013-03-18 2017-04-04 Spotify Ab Cross media recommendation
US20200159744A1 (en) * 2013-03-18 2020-05-21 Spotify Ab Cross media recommendation
US11645301B2 (en) 2013-03-18 2023-05-09 Spotify Ab Cross media recommendation
WO2014153133A1 (en) * 2013-03-18 2014-09-25 The Echo Nest Corporation Cross media recommendation
US10558682B2 (en) * 2013-03-18 2020-02-11 Spotify Ab Cross media recommendation
US20140279756A1 (en) * 2013-03-18 2014-09-18 The Echo Nest Corporation Cross media recommendation
US10339586B1 (en) * 2013-08-23 2019-07-02 Amazon Technologies, Inc. Techniques for identifying similar products
US10250933B2 (en) 2013-10-08 2019-04-02 Spotify Ab Remote device activity and source metadata processor
US9451329B2 (en) 2013-10-08 2016-09-20 Spotify Ab Systems, methods, and computer program products for providing contextually-aware video recommendation
WO2015054361A1 (en) * 2013-10-08 2015-04-16 The Echo Nest Corporation Systems, methods, and computer program products for providing contextually-aware video recommendation
WO2016207308A1 (en) * 2015-06-26 2016-12-29 British Telecommunications Public Limited Company Digital content provision
US10893312B2 (en) 2015-06-26 2021-01-12 British Telecommunications Public Limited Company Digital content provision
US10516906B2 (en) 2015-09-18 2019-12-24 Spotify Ab Systems, methods, and computer products for recommending media suitable for a designated style of use
US12001500B2 (en) 2015-11-17 2024-06-04 Spotify Ab System, methods and computer products for determining affinity to a content creator
US9798823B2 (en) 2015-11-17 2017-10-24 Spotify Ab System, methods and computer products for determining affinity to a content creator
US11210355B2 (en) 2015-11-17 2021-12-28 Spotify Ab System, methods and computer products for determining affinity to a content creator
EP3203756A1 (de) * 2016-02-02 2017-08-09 Thomson Licensing Verfahren zur bestimmung eines personalisierten profils und zum filtern von auszügen eines multimediainhalts, zugehörige vorrichtungen, computerprogrammprodukt und computerlesbares trägermedium
US11537657B2 (en) 2016-08-18 2022-12-27 Spotify Ab Systems, methods, and computer-readable products for track selection
US10860646B2 (en) 2016-08-18 2020-12-08 Spotify Ab Systems, methods, and computer-readable products for track selection
CN110110209A (zh) * 2018-01-22 2019-08-09 青岛科技大学 一种基于局部加权线性回归模型的交叉推荐方法及系统
GB2606878A (en) * 2018-05-25 2022-11-23 Thinkanalytics Ltd Content recommendation system
GB2606878B (en) * 2018-05-25 2023-02-22 Thinkanalytics Ltd Content recommendation system
US11812107B2 (en) 2018-05-25 2023-11-07 Thinkanalytics Ltd Content recommendation system and method
WO2020096272A1 (en) * 2018-11-05 2020-05-14 Samsung Electronics Co., Ltd. System and method for cross-domain recommendations
US11604844B2 (en) 2018-11-05 2023-03-14 Samsung Electronics Co., Ltd. System and method for cross-domain recommendations
EP3902277A4 (de) * 2019-01-07 2022-06-08 Samsung Electronics Co., Ltd. Verfahren zur bereitstellung einer liste mit empfohlenen inhalten und entsprechende elektronische vorrichtung
US11379513B2 (en) 2019-08-30 2022-07-05 Spotify Ab Systems and methods for providing content based on consumption in a distinct domain
US11681739B2 (en) 2019-08-30 2023-06-20 Spotify Ab Systems and methods for providing content based on consumption in a distinct domain
CN111046280B (zh) * 2019-12-02 2023-12-12 哈尔滨工程大学 一种应用fm的跨领域推荐方法
CN111046280A (zh) * 2019-12-02 2020-04-21 哈尔滨工程大学 一种应用fm的跨领域推荐方法
JPWO2021176716A1 (de) * 2020-03-06 2021-09-10
WO2021176716A1 (ja) * 2020-03-06 2021-09-10 日本電気株式会社 嗜好推定装置、嗜好推定方法および嗜好推定プログラム
US12114030B1 (en) 2023-04-04 2024-10-08 Thinkanalytics Ltd. Managing feature distribution

Also Published As

Publication number Publication date
EP2207348A3 (de) 2013-02-13

Similar Documents

Publication Publication Date Title
EP2207348A2 (de) Empfehlungsverfahren und System zur Cross-Domain-Empfehlung
CN103052954B (zh) 推荐系统的基于简档内容检索
CA3041380C (en) Recommendation engine apparatus and methods
JP5032477B2 (ja) ユーザに関心を引くアイテムを推奨するシステムおよび方法
JP2023105206A (ja) 会話におけるコンテキスト切替を決定するためのシステムおよび方法
US11100157B2 (en) Determining recommended media programming from sparse consumption data
US20100011020A1 (en) Recommender system
US20140013353A1 (en) Social Network Based Recommendation Method and System
KR20020033202A (ko) 3가지 방식의 매체 추천 방법 및 시스템 명세
EP2382560A1 (de) Adaptives implizites lernen für empfehlungsvorrichtungssysteme
KR102147458B1 (ko) 시청 패턴에 기초한 맞춤형 콘텐츠 및 광고 제공 시스템 및 방법
US20110041157A1 (en) Systems and Methods for Selecting Content For a Subscriber of a Content Service Provider
US9445152B2 (en) Recommender system with consistent profile application
US11064260B2 (en) System and method for locating content related to a media asset
US8700556B2 (en) Biased recommender system based on age parameter
US20210160591A1 (en) Creating customized short-form content from long-form content
Pogacnik et al. Personal content recommender based on a hierarchical user model for the selection of TV programmes
Mukherjee et al. A context-aware recommendation system considering both user preferences and learned behavior
Adeliyi et al. Minimizing zapping delay using adaptive channel switching with personalized electronic program guide
KR101223137B1 (ko) 수요자 기반 메타데이터를 이용한 창작 지원 시스템 및 이의 정보 제공 방법
Zhu et al. Content classification and recommendation techniques for viewing electronic programming guide on a portable device

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO SE SI SK TR

AX Request for extension of the european patent

Extension state: AL BA RS

PUAL Search report despatched

Free format text: ORIGINAL CODE: 0009013

RIC1 Information provided on ipc code assigned before grant

Ipc: H04N 5/445 20110101AFI20121219BHEP

Ipc: G06Q 30/00 20120101ALI20121219BHEP

Ipc: G06F 17/30 20060101ALI20121219BHEP

AK Designated contracting states

Kind code of ref document: A3

Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO SE SI SK TR

AX Request for extension of the european patent

Extension state: AL BA RS

17P Request for examination filed

Effective date: 20130813

RBV Designated contracting states (corrected)

Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO SE SI SK TR

AKX Designation fees paid

Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO SE SI SK TR

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: FUNKE DIGITAL TV GUIDE GMBH

17Q First examination report despatched

Effective date: 20180511

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION HAS BEEN REFUSED

18R Application refused

Effective date: 20191108